{"paper_id":"2611.19792","title":"PIRanNN: Random-Feature Pricing and Calibration under Jump-Diffusion and Stochastic-Volatility Models","authors":["Saurabh Bansal","Srikanth K. Iyer","Shashi Jain"],"paper_date":"2026-11-15","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2611.19792","page_url":"https://thequant.space/flowcharts/pirannn-random-feature-pricing-and-calibration-under-jump-diffusion-and/","scores":{"empirical_rigor":7,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a random-feature method for pricing and calibrating European options under jump‑diffusion (Merton, Kou) and stochastic‑volatility (Heston, SABR) models.","quote":"We study a random-feature method for pricing and calibrating European options under jump-diffusion (Merton, Kou) and stochastic-volatility (Heston, SABR) models.","summary":"We study a random-feature method for pricing and calibrating European options under jump-diffusion (Merton, Kou) and stochastic-volatility (Heston, SABR) models. The network has one hidden layer with a probit activation and fixed random inner weights, so fitting it to the pricing equation, the…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The interest rate r is constant.","type":"model","stated":"explicit","quote":"We consider a European derivative on this asset that pays g(ST ) at maturity T, for a given payoff function g, and denote by V (S, t) its price at time t when St = S. For ease of notation, we take the interest rate r to be constant, so the bank account is Bt = ert.","support":"full"},{"assumption":"Pricing is performed under the risk‑neutral measure Q.","type":"model","stated":"explicit","quote":"For the underlying, this means that e−(r−q)tSt is a martingale, and the price of the derivative is the expected discounted payoff, V (S, t) = EQ \u0002 e−r(T −t)g(ST ) St = S \u0003 .","support":"full"},{"assumption":"The jump integral of each random feature can be evaluated exactly for Merton and Kou models.","type":"model","stated":"explicit","quote":"The jump term is exact. Averaging a probit feature over a normal or a double-exponential jump gives an explicit formula (Lemma 4.3).","support":"full"},{"assumption":"A Black‑Scholes price subtraction is applied to remove the payoff kink.","type":"other","stated":"implied","quote":"3. For the jump models, a quadrature-free Black–Scholes subtraction removes the main source of error.","support":"full"},{"assumption":"The underlying asset may pay a continuous dividend yield q ≥ 0.","type":"model","stated":"explicit","quote":"Let St be the price at time t of the underlying asset, which pays a continuous dividend yield q ≥0 (for a currency, q is the foreign interest rate).","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Synthetic test problems for Merton, Kou, Heston and SABR pricing and calibration","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"Experiment Model parameters Domain m nens Merton pricing r = 0.05, T = 1, σ = 0.2, λ = 0.1, µJ = −0.1, σJ = 0.3 x ∈[−1.2, 1.2], τ ∈[0, 1]; finite differences on [−2.5, 2.5] 1800 4 Kou pricing r = 0.05, T = 1, σ = 0.2, λ = 0.3, p = 0.4, η1 = 10, η2 = 5 x ∈[−1.2, 1.2], τ ∈[0, 1]; finite differences on [−2.5, 2.5] 1800 4 Jump calibration as for pricing x ∈[−1.2, 1.2], τ ∈[0, 1] 800 1 (5 in the repeat study) PINN","support":"full"},{"name":"Monte Carlo simulation reference for SABR","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"On a separate grid with maturities up to five years and strikes up to K/F0 = 2, compared with Monte Carlo simulation of (3.8) (400,000 paths, 500 time steps), PIRanNN is at least as close to the Monte Carlo price as Hagan’s formula at 9 of the 10 points. We simulate α exactly on a time grid, with 8 million paths, and the standard error is below 0.01% of the price at every quote.","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"pricing-calibration","findings":[{"finding":"The method matches the accuracy of a recent physics‑informed neural network in seconds instead of about half an hour.","kind":"performance","quote":"On the test problems of a recent physics-informed neural network study, the method matches that network's accuracy in seconds instead of about half an hour.","support":"full"},{"finding":"Subtracting the Black‑Scholes price reduces the pricing error for jump models by roughly two orders of magnitude.","kind":"pricing-calibration","quote":"For the jump models, subtracting the Black-Scholes price removes the payoff kink and reduces the pricing error by about two orders of magnitude.","support":"full"},{"finding":"Without the subtraction the calibration misses jump parameters by up to 50 %; with it the parameters are recovered within 3 %.","kind":"pricing-calibration","quote":"The calibration experiments show that a solver can fit the quotes and still return wrong parameters when its pricing error is not small compared with the effect of the parameters on the prices: without the subtraction the solver misses the jump parameters by up to 50%, and with it they are recovered to within 3%.","support":"full"},{"finding":"For Heston and SABR the pricing errors are about one percent and calibrated parameters are within about 15 % of the true values.","kind":"pricing-calibration","quote":"For Heston and SABR, used as an applicability study with boundary conditions that need no reference prices, pricing errors are about one percent and the calibrated parameters are within about 15% of the true values.","support":"full"},{"finding":"PIRanNN with subtraction attains an RMSE of 4.5 × 10⁻⁷ for Merton and 1.9 × 10⁻⁶ for Kou, comparable to a Crank‑Nicolson finite‑difference scheme.","kind":"pricing-calibration","quote":"PIRanNN with subtraction is as accurate for Merton (4.5 × 10−7) and about four times less accurate for Kou (1.9 × 10−6).","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":36,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":true,"code_url":"https://github.com/RiskLabIISc/PiRANN","note":null},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":3,"extracted_at":"2026-11-17T15:29:23Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}